Preemptive Toxic Language Detection in Wikipedia Comments Using Thread-Level Context

Karan, Mladen; Šnajder, Jan

doi:10.18653/v1/w19-3514

Cited by 28 publications

(25 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is, however, evident from our empirical results that incorporating context through providing the preceding comment to the model did not improve performance for both traditional machine learning and deep learning models. This finding is consistent with other studies that attempt to incorporate interactional context into their models (Karan and Šnajder, 2019;. We believe that effectively incorporating deeper context, as opposed to just the preceding comment, using more sophisticated methods such as hierarchical neural networks might help improve performance.…”

Section: Contextsupporting

confidence: 90%

“Are you kidding me?”: Detecting Unpalatable Questions on Reddit

Bagga¹,

Piper

Ruths

2021

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume

View full text Add to dashboard Cite

Abusive language in online discourse negatively affects a large number of social media users. Many computational methods have been proposed to address this issue of online abuse. The existing work, however, tends to focus on detecting the more explicit forms of abuse leaving the subtler forms of abuse largely untouched. Our work addresses this gap by making three core contributions. First, inspired by the theory of impoliteness, we propose a novel task of detecting a subtler form of abuse, namely unpalatable questions. Second, we publish a context-aware dataset for the task using data from a diverse set of Reddit communities. Third, we implement a wide array of learning models and also investigate the benefits of incorporating conversational context into computational models. Our results show that modeling subtle abuse is feasible but difficult due to the language involved being highly nuanced and context-sensitive. We hope that future research in the field will address such subtle forms of abuse since their harm currently passes unnoticed through existing detection systems.

show abstract

Section: Contextsupporting

confidence: 90%

“Are you kidding me?”: Detecting Unpalatable Questions on Reddit

Bagga¹,

Piper

Ruths

2021

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume

View full text Add to dashboard Cite

show abstract

“…The authors showed that certain features in the first messages of a conversation, such as the use of first or second person pronouns and the presence of certain politeness strategies, can help predict if that conversation will remain healthy or if it will degrade and lead to harmful messages later on. Their work inspired the authors of [13], who trained and tested an SVM using TFIDF-weighted unigrams and bigrams as well as a BiLSTM using their own word embeddings. They were ultimately dissatisfied with their results, however the fact they focused only on words and did not use more sophisticated features such as those in [1] may be the cause.…”

Section: Related Workmentioning

confidence: 99%

Using Sentiment Information for Preemptive Detection of Harmful Comments in Online Conversations

Brassard-Gourdeau¹,

Khoury

2021

Proceedings of the Canadian Conference on Artificial Intelligence

View full text Add to dashboard Cite

The challenge of automatic moderation of harmful comments online has been the subject of a lot of research recently, but the focus has been mostly on detecting it in individual messages after they have been posted. Some authors have tried to predict if a conversation will derail into harmfulness using the features of the first few messages [1]. In this paper, we combine that approach with previous work on harmful message detection using sentiment information [2], and show how the sentiments expressed in the first messages of a conversation can help predict upcoming harmful messages. Our results show that adding sentiment features does help improve the accuracy of harmful message prediction, and allow us to make important observations on the general task of preemptive harmfulness detection.

show abstract

“…Recent work of volunteer moderators and moderation mainly focuses on user-governed platforms such as Wikipedia [11,33], and Reddit [9,16,29]. Twitch, as a user-moderated, live-streaming community, is similar in some governance aspects to other online communities such as Reddit, which is a self-reliant community [29], and Facebook Group, which provides multiparty interactions [51].…”

Section: Introductionmentioning

confidence: 99%

Moderation Visibility: Mapping the Strategies of Volunteer Moderators in Live Streaming Micro Communities

Cai

Wohn

Almoqbel

2021

ACM International Conference on Interactive Media Experiences

View full text Add to dashboard Cite

Volunteer moderators actively engage in online content management, such as removing toxic content and sanctioning anti-normative behaviors in user-governed communities. The synchronicity and ephemerality of live-streaming communities pose unique moderation challenges. Based on interviews with 21 volunteer moderators on Twitch, we mapped out 13 moderation strategies and presented them in relation to the bad act, enabling us to categorize from proactive and reactive perspectives and identify communicative and technical interventions. We found that the act of moderation involves highly visible and performative activities in the chat and invisible activities involving coordination and sanction. The juxtaposition of real-time individual decision-making with collaborative discussions and the dual nature of visible and invisible activities of moderators provide a unique lens into a role that relies heavily on both the social and technical. We also discuss how the afordances of live-streaming contribute to these unique activities. CCS CONCEPTS• Human-centered computing → Empirical studies in collaborative and social computing; Empirical studies in HCI .

show abstract

Preemptive Toxic Language Detection in Wikipedia Comments Using Thread-Level Context

Cited by 28 publications

References 22 publications

“Are you kidding me?”: Detecting Unpalatable Questions on Reddit

“Are you kidding me?”: Detecting Unpalatable Questions on Reddit

Using Sentiment Information for Preemptive Detection of Harmful Comments in Online Conversations

Moderation Visibility: Mapping the Strategies of Volunteer Moderators in Live Streaming Micro Communities

Contact Info

Product

Resources

About